Table 3. Proportion of spectral Solar irradiance at surface by spectral 
region. Also given is percent diffuse to global. Analysis of Dave’ 
(1978) atmosphere diffuse data set. Model 2 Gases with no aerosols. 
Model 3 gases plus low loading (r’ a =0.09 0 5 fjm) . Model 4 gases plus 
moderate aerosol loading (r’ a =0.45 0 .5 /¿m) . UV (0.31-0.38/im) , VS(.38- 
.73um), Ml(.72-1 3|Uin) and 141(1.3-2.41/an) . 
Sun Zenith Angle 
Model 2 
0° 
UV VS Mi MR 
UV 
Direct 
.039 .471 .385 .105 
.031 
Diffuse 
.313 .635 .051 .001 
.288 
Global 
.053 .479 .368 .100 
.048 
Diff(%) 
29.6 6.6 0.7 0.1 
39.8 
Model 3 
Direct 
.038 .466 .386 .110 
.031 
Diffuse 
.154 . 568 . 242 . 034 
.142 
Global 
.052 .478 .370 .101 
.048 
Diff (55) 
34.4 13.8 7.6 3.9 
45.6 
Model 4 
Direct 
.037 .444 .390 .129 
.029 
Diffuse 
.077 . 528 . 340 . 056 
.070 
Global 
.050 .472 .373 .104 
.046 
Diff (%) 
50.8 37.4 30.5 17.8 
63.9 
45o 7oo 
VS 
Mi 
Mi 
UV 
VS 
NR 
Mi 
.468 
.395 
.106 
.014 
.474 
.431 
.110 
.658 
.054 
.001 
.220 
.711 
.850 
.001 
480 
.373 
.098 
.039 
.460 
.387 
.097 
9.2 
1.0 
0.1 
67.6 
18.0 
2.4 
0.2 
460 
.397 
.112 
.014 
.428 
.434 .124 
579 
.245 
.033 
.111 
.600 
.257 .033 
478 
.374 .099 
.040 .474 
.386 .099 
L8.7 
10.1 
5.1 
74.3 33.8 
17.8 8.8 
.429 
.402 
.140 
.012 
.360 
.434 
.194 
.528 
.345 
.056 
.058 
.521 
.359 
.062 
.471 
.378 
.105 
.042 
.464 
.385 
.108 
47.5 
38.6 
22.7 
89.9 
72.8 
60.4 
37.1 
high aerosol loading) at selected Sun zenith 
angles was used to examine the spectral region 
variability. The global radiation percentages 
given in Table 3 may be used to represent the 
spectral region solar irradiance coefficients 
(i.e., k) in Equation (6). As expected for both 
an increasing Sun zenith angle (i. e. , CP to 7(P 
6 S ) and an increasing aerosol loading (Model 2 to 
3 to 4), there is increased scattering, producing 
less direct radiation and more diffuse radiation. 
For direct radiation, there is proportionally 
less energy in the ultraviolet and visible 
regions with a shift of energy to the near-IR and 
middle-IR wavelengths. However, as evident by 
the small changes in the global radiation 
proportions between illumination and atmospheric 
haze conditions, there is a strong compensating 
effect of an increased diffuse radiation from 
forward Mie scattering adding to the reduced 
direct radiation. Differences between values in 
the visible, near-IR and middle-IR have a range 
of less than 3%. Hence, examination of results 
indicates a single set of spectral region 
proportions for the global radiation may be used 
in Equation (6). The spectral proportions used 
were from the 45° Sian zenith angle for the heavy 
aerosol model (Model-4) of 0.046 ultraviolet, 
0.471 visible, 0.378 near-IR, and 0.105 middle- 
IR. 
3.5 Surface Temperature Estimation 
The calibrated brightness temperature for 
bands 4 and 5 (Kidwell 1983) are used in a split 
window technique to correct for atmospheric 
absorption effects (Strong and McClain 1984) . 
The technique was developed to extract sea 
surface temperatures, assumirjg a constant sea- 
surface emissivity. 
17 = 3.6125*TB-11 - 2.5779*TB-12 - 10.05 (7) 
Tg = 3.6446*Tb~11 - 2.6616*TB-12 + 5.2 (8) 
T - surface temperature corrected for 
atmospheric effects for NOAA-7 and 
NOAA-9 respectively 
TB-!! - AVHRR band 4 brightness temperature 
T}3-12 _ AVHRR band 5 brightness temperature 
The root mean square of AVHRR derived surface 
temperature versus drifting buoy readings is 
reported at 0.6°C (Strong and McClain 1984). 
Unfortunately, small differences between the sea- 
surface emissivity and the ground emissivity may 
lead to large derived ground temperature errors. 
For land surfaces, Cooper and Asrar (1989) 
indicate the NOAA split window technique provides 
land surface temperatures within + 3°C of In situ 
measurements. To adjust the split window derived 
soar face temperatures for land surfaces with an 
emissivity different than that for water, the 
following relationship from Vukovich et al. 
(1987) was used. 
Tg = T c (e s /6g)l/n 
Tg - ground temperature 
e s - sea surface emissivity 
€g - ground emissivity 
n — temperature exponent (I = eT 11 ) 
T c - temperature frcm split window procedure 
The value of n, according to Price (1985), is 
approximately 4.5. A sea surface emissivity (e s ) 
of 0.99 was used. The ground emissivity (eg) 
was estimated for the study site to range frcm 
0.93 in the arid region to the north to 0.97 in 
the more densely vegetated area to the south 
(Wolfe and Zissis 1978 and Taylor 1979). The 
emissivity is approximately proportional to the 
amount of vegetation and soar face water. We 
assumed the derived spectral vegetation index, 
ranging between 0 and 0.6 was linearly related to 
the surface emissivity. Hence, a ground 
emissivity was scaled by the NDVI using the 
equation given next. 
€g = 0.93 + (NDVI * 0.0667) (9) 
4. ANALYSIS REPORT 
Monthly mean precipitation, surface air and 
dew-point temperature, and the height of the 
lifting condensation level (LCL) relative to the 
surface are given in Figures 2 and 3 for 
Tambaoounda and Podor. Analysis of weather data 
for Podor and Tambarounda indicates the rainy 
seasons occur during midsummer and are associated 
with the migration of the intertropical 
convergence zone and the accompanying horizontal 
flux of moisture in the lower levels. The 
initiation of the rainy season has occurred as 
early as April and terminated as late as October. 
Though the amount of atmospheric moisture is 
about the same at the two locations, the rainfall 
is greater at Tambaoounda, which is south of 
Podor. At both locations, the influx of moisture 
produces a lower LCL height which increases the 
potential for convective clouds and 
precipitation, since less work will be required, 
under this condition, to produce clouds and 
precipitation. At Tambaoounda, the LCL height is 
lower than that at Podor apparently due to the 
lower surface air temperature, which is brought 
about by the greater cloudiness at Tambaoounda 
(i.e. , greater precipitation amounts) since the 
moisture is about the same. This factor may be 
attributed to a feedback mechanism since the 
potential for cloudiness and precipitation 
becomes greater at Tambaoounda, with a lower LCL 
height. 
Tables 4 and 5 provides the resultant AVHRR 
derived surface parameters for the case studies 
over the period 1981 to 1985. Derived parameters 
presented in the table represent a 10 x 10 pixel 
area average. The standard deviation over the 
area is also presented in Tables 4 and 5 and 
these are the values in the parentheses. Tables 
4 and 5 also contain values representing two to 
four week antecedent precipitation amount. It 
260